Simpler stz manager #8
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Mesa DescriptionOverviewThis PR refactors the
TestingRan locally + against a few unikernels in staging. I would connect and disconnect a bunch of tabs to simulate flaky connections (this was a pretty consistent repro of the original issue) Example logs from staging: Description generated by Mesa. Update settings |
![mesa-dot-dev[bot]](https://avatars.githubusercontent.com/in/1050077?s=60&v=4){kind=small}
There was a problem hiding this comment.
Performed full review of 46e2fc7...e9293ae
Analysis
-
Despite the architectural improvements, the PR doesn't address potential race conditions in the new state-based design that could occur during concurrent session state transitions.
-
The simplified boolean flag
disabledScaleToZeromay lack granularity compared to the counter approach, potentially making it difficult to handle partial or in-progress scaling operations. -
The centralized
manage()method creates a potential single point of failure and may become a bottleneck as system complexity grows. -
The PR appears to focus on fixing symptoms rather than addressing the root cause of event deduplication failures, which could resurface in other components.
Tip
⚡ Quick Actions
This review was generated by Mesa.
Actions:
Slash Commands:
/review- Request a full code review/review latest- Review only changes since the last review/describe- Generate PR description. This will update the PR body or issue comment depending on your configuration/help- Get help with Mesa commands and configuration options
1 files reviewed | 0 comments | Review on Mesa | Edit Reviewer Settings
rgarcia
approved these changes
Oct 21, 2025
| connectedSessions := 0 | ||
| for _, s := range m.sessions.List() { | ||
| if s.State().IsConnected { | ||
| connectedSessions++ |
There was a problem hiding this comment.
is there a world where a session is disconnected (IsConnected == false) but is trying to reconnect, and by not counting it here we might scale to zero, preventing it from connecting?
Collaborator
Author
There was a problem hiding this comment.
Great question! There's a grace period after disconnects for reconnection: https://github.com/onkernel/neko/blob/e9293ae098832ae0c4b63d092a9bd370e3c0ab12/server/internal/session/session.go#L145-L213
Especially wrt to scale to zero that delay + this buffer seems reasonable. For new connections at our proxy layer we'd be "waking up" the VM externally regardless ^^
Sayan-
added a commit
to kernel/kernel-images
that referenced
this pull request
Oct 22, 2025
Pulls in kernel/neko#8 kernel/neko#9 <!-- CURSOR_SUMMARY --> --- > [!NOTE] > Update chromium-headful Dockerfile to use neko base image 3.0.8-v1.3.0. > > - **Docker**: > - Bump base image in `images/chromium-headful/Dockerfile` from `ghcr.io/onkernel/neko/base:3.0.8-v1.1.0` to `3.0.8-v1.3.0`. > > <sup>Written by [Cursor Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit e6afb32. This will update automatically on new commits. Configure [here](https://cursor.com/dashboard?tab=bugbot).</sup> <!-- /CURSOR_SUMMARY -->
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Overview
We noticed issues where implementation would have < 0 pending. It seems like the naive assumptions that events would be deduped and delivered exactly once weren't really reliable. Instead swap to a model where we only disable/enable around edges rather than every event
Testing
Ran locally + against a few unikernels in staging. I would connect and disconnect a bunch of tabs to simulate flaky connections (this was a pretty consistent repro of the original issue)
Example logs from staging: